基于笔划趋势分析的二值图像细化方法

Thinning of binary image based on stroke trend analysis

针对普通细化算法大多是基于原始图的中轴线及八邻域进行分析,不可避免导致有噪声点的直线变弯以及笔划交叉或分叉位置产生畸变的问题,在对Zhang算法进行分析和改进的基础上,提出了一种改进的基于笔划趋势预测的细化算法,获得了比较严格八连通的且各向对称的中心骨架,并有效地抑制了笔划交叉处的畸变,具有很好的细化效果和较快的处理速度,已成功运用于车牌自动识别系统中。

In order to overcome the drawbacks of thinning based on the axle wire of the original image and eight-neighborhood, which causes some straight-lines with noise to become curves, and stroke distortion at the cross or crotch of strokes, an improved thinning algorithm based on stroke trend analysis is proposed for binary image by analyzing and improving Zhang thinning algorithm, and it can make a skeleton with strictly eight-connected structure available to keep a good symmetry, and control the deformation at the cross or crotch of strakes so that each image can be quickly skeletonized more perfectly.The algorithm proposed in this paper has been successfully applied to vehicle license identifying system.